The Framingham Study has been abundantly reported in the literature and will be outlined here only briefly. It was established in 1948-1949 by the National Heart Institute to follow 2282 men and 2845 women, 22-62 years of age, who were initially free of coronary heart disease. Each subject at baseline received an extensive standardized cardiovascular examination, which included information on habits, physical characteristics, and blood chemistries. Biennially thereafter, and now extending for more than 40 years, the subjects have been thoroughly re-examined for the same initial characteristics, with development of CHD [coronary heart disease] noted as the ultimate endpoint.
Things got interesting when Seltzer noticed that moderate drinkers had lower rates of cardiovascular disease than teetotallers. This was 45 years ago.
Early in 1972, Dr. William Kannel, who was then assistant director of the Framingham Study, gave me a copy of the Framingham IBM cards containing data for all the participants of the Study. My assignment was to analyze the data with respect to such factors as obesity and body build. During the analysis, I found that after 16 years of follow-up, the Framingham men who habitually consumed moderate amounts of alcohol showed a lesser risk of developing CHD than those who refrained from alcohol drinking. The risk of CHD also showed a “dose-response curve,” i.e., a regularly diminishing gradient proportional to the amount of alcohol consumed. To my knowledge, this result was the first time this phenomenon had been observed in CHD research.
Seltzer wrote up his findings in 1972 and was told to send his paper to the National Heart and Lung Institute.
I sent the revised manuscript on June 26 to Dr. William J. Zukel, then Associate Director for Clinical Applications, NHLI. Dr. Zukel never directly answered my letter. Instead, I next heard from Dr. Kannel, enclosing a memorandum he had received from Dr. Zukel, who refused to allow the alcohol/CHD manuscript to be submitted for publication. The main stated reason was that “An article which openly invites the encouragement of undertaking drinking with the implication of prevention of coronary heart disease would be scientifically misleading and socially undesirable in view of the major health problem of alcoholism that already exists in the country”. Dr. Zukel added that it would not be appropriate “to have such a manuscript with these unsupportable conclusions co-authored from the staff of the NHLI."He urged that an article be produced maintaining the “conclusion of no significant relationship of alcohol intake to incidence of coronary heart disease”. The results would be based on earlier data showing no association between alcohol consumption and CHD. Dr. Zukel also questioned the statistical significance of findings in the submitted manuscript, and asked that I return the punchcards given to me by Dr. Kannel. Since the alcohol data were not my own, and since Dr. Zukel had prohibited publication, I had no choice but to regard the manuscript as defunct.
And so the evidence was buried because it might send out the wrong message. Writing in 1997, Seltzer treats the episode as an example of bias in government-funded research, but seems to hold no grudge since the association between moderate drinking and better heart health had by then been established beyond doubt.
Despite Dr. Zukel’s suppression of the original scientific results, the Framingham finding about alcohol and CHD has been validated in other studies, and has now become the conventional view in the CHD literature.
Little did he know that a new generation of anti-alcohol researchers like Tim Stockwell (who is now calling for the nationalisation of the alcohol industry) would launch a renewed politically-motivated assault on the evidence.
The main point of this essay, however, is that conflicts of interest and pressures on investigators need not arise exclusively from commercial organizations. A non-profit governmental agency that funds research can also suppress some of its findings, and can alter definitions and analyses to make results that originally contradict a governmental policy emerge as supportive. What should be epidemiologic science may then become political science.
This is a highly perceptive and prescient comment given Public Health England's behaviour in recent years. Some things never change.